- Single header file
- Requires C++17
- MIT License
Simply include argparse.hpp and you're good to go.
#include <argparse/argparse.hpp>To start parsing command-line arguments, create an ArgumentParser.
argparse::ArgumentParser program("program_name");NOTE: There is an optional second argument to the ArgumentParser which is the program version. Example: argparse::ArgumentParser program("libfoo", "1.9.0");
To add a new argument, simply call .add_argument(...). You can provide a variadic list of argument names that you want to group together, e.g., -v and --verbose
program.add_argument("foo");
program.add_argument("-v", "--verbose"); // parameter packingArgparse supports a variety of argument types including positional, optional, and compound arguments. Below you can see how to configure each of these types:
Here's an example of a positional argument:
#include <argparse/argparse.hpp>
int main(int argc, char *argv[]) {
argparse::ArgumentParser program("program_name");
program.add_argument("square")
.help("display the square of a given integer")
.scan<'i', int>();
try {
program.parse_args(argc, argv);
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
auto input = program.get<int>("square");
std::cout << (input * input) << std::endl;
return 0;
}And running the code:
$ ./main 15
225Here's what's happening:
- The
add_argument()method is used to specify which command-line options the program is willing to accept. In this case, I’ve named it square so that it’s in line with its function. - Command-line arguments are strings. To square the argument and print the result, we need to convert this argument to a number. In order to do this, we use the
.scanmethod to convert user input into an integer. - We can get the value stored by the parser for a given argument using
parser.get<T>(key)method.
Now, let's look at optional arguments. Optional arguments start with - or --, e.g., --verbose or -a. Optional arguments can be placed anywhere in the input sequence.
argparse::ArgumentParser program("test");
program.add_argument("--verbose")
.help("increase output verbosity")
.default_value(false)
.implicit_value(true);
try {
program.parse_args(argc, argv);
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
if (program["--verbose"] == true) {
std::cout << "Verbosity enabled" << std::endl;
}$ ./main --verbose
Verbosity enabledHere's what's happening:
- The program is written so as to display something when --verbose is specified and display nothing when not.
- Since the argument is actually optional, no error is thrown when running the program without
--verbose. Note that by using.default_value(false), if the optional argument isn’t used, it's value is automatically set to false. - By using
.implicit_value(true), the user specifies that this option is more of a flag than something that requires a value. When the user provides the --verbose option, it's value is set to true.
There are scenarios where you would like to make an optional argument required. As discussed above, optional arguments either begin with - or --. You can make these types of arguments required like so:
program.add_argument("-o", "--output")
.required()
.help("specify the output file.");If the user does not provide a value for this parameter, an exception is thrown.
Alternatively, you could provide a default value like so:
program.add_argument("-o", "--output")
.default_value(std::string("-"))
.required()
.help("specify the output file.");If you require an optional argument to be present but have no good default value for it, you can combine testing and accessing the argument as following:
if (auto fn = program.present("-o")) {
do_something_with(*fn);
}Similar to get, the present method also accepts a template argument. But rather than returning T, parser.present<T>(key) returns std::optional<T>, so that when the user does not provide a value to this parameter, the return value compares equal to std::nullopt.
If you want to know whether the user supplied a value for an argument that has a .default_value, check whether the argument .is_used().
program.add_argument("--color")
.default_value(std::string{"orange"}) // might otherwise be type const char* leading to an error when trying program.get<std::string>
.help("specify the cat's fur color");
try {
program.parse_args(argc, argv); // Example: ./main --color orange
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
auto color = program.get<std::string>("--color"); // "orange"
auto explicit_color = program.is_used("--color"); // true, user provided orangeYou may want to allow an optional argument to be repeated and gather all values in one place.
program.add_argument("--color")
.default_value<std::vector<std::string>>({ "orange" })
.append()
.help("specify the cat's fur color");
try {
program.parse_args(argc, argv); // Example: ./main --color red --color green --color blue
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
auto colors = program.get<std::vector<std::string>>("--color"); // {"red", "green", "blue"}Notice that .default_value is given an explicit template parameter to match the type you want to .get.
A common pattern is to repeat an argument to indicate a greater value.
int verbosity = 0;
program.add_argument("-V", "--verbose")
.action([&](const auto &) { ++verbosity; })
.append()
.default_value(false)
.implicit_value(true)
.nargs(0);
program.parse_args(argc, argv); // Example: ./main -VVVV
std::cout << "verbose level: " << verbosity << std::endl; // verbose level: 4Optional arguments start with -. Can argparse handle negative numbers? The answer is yes!
argparse::ArgumentParser program;
program.add_argument("integer")
.help("Input number")
.scan<'i', int>();
program.add_argument("floats")
.help("Vector of floats")
.nargs(4)
.scan<'g', float>();
try {
program.parse_args(argc, argv);
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
// Some code to print arguments$ ./main -5 -1.1 -3.1415 -3.1e2 -4.51329E3
integer : -5
floats : -1.1 -3.1415 -310 -4513.29As you can see here, argparse supports negative integers, negative floats and scientific notation.
argparse::ArgumentParser program("main");
program.add_argument("square")
.help("display the square of a given number")
.scan<'i', int>();
program.add_argument("--verbose")
.default_value(false)
.implicit_value(true);
try {
program.parse_args(argc, argv);
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
int input = program.get<int>("square");
if (program["--verbose"] == true) {
std::cout << "The square of " << input << " is " << (input * input) << std::endl;
}
else {
std::cout << (input * input) << std::endl;
}$ ./main 4
16
$ ./main 4 --verbose
The square of 4 is 16
$ ./main --verbose 4
The square of 4 is 16std::cout << program prints a help message, including the program usage and information about the arguments registered with the ArgumentParser. For the previous example, here's the default help message:
$ ./main --help
Usage: main [options] square
Positional arguments:
square display the square of a given number
Optional arguments:
-h --help shows help message and exits [default: false]
-v --version prints version information and exits [default: false]
--verbose [default: false]
You may also get the help message in string via program.help().str().
ArgumentParser::add_description will add text before the detailed argument
information. ArgumentParser::add_epilog will add text after all other help output.
argparse::ArgumentParser program("main");
program.add_argument("thing")
.help("Thing to use.");
program.add_description("Forward a thing to the next member.");
program.add_epilog("Possible things include betingalw, chiz, and res.");
program.parse_args(argc, argv);
std::cout << program << std::endl;$ ./main --help
Usage: main thing
Forward a thing to the next member.
Positional arguments:
thing Thing to use.
Optional arguments:
-h --help shows help message and exits [default: false]
-v --version prints version information and exits [default: false]
Possible things include betingalw, chiz, and res.ArgumentParser objects usually associate a single command-line argument with a single action to be taken. The .nargs associates a different number of command-line arguments with a single action. When using nargs(N), N arguments from the command line will be gathered together into a list.
argparse::ArgumentParser program("main");
program.add_argument("--input_files")
.help("The list of input files")
.nargs(2);
try {
program.parse_args(argc, argv); // Example: ./main --input_files config.yml System.xml
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
auto files = program.get<std::vector<std::string>>("--input_files"); // {"config.yml", "System.xml"}ArgumentParser.get<T>() has specializations for std::vector and std::list. So, the following variant, .get<std::list>, will also work.
auto files = program.get<std::list<std::string>>("--input_files"); // {"config.yml", "System.xml"}Using .scan, one can quickly build a list of desired value types from command line arguments. Here's an example:
argparse::ArgumentParser program("main");
program.add_argument("--query_point")
.help("3D query point")
.nargs(3)
.default_value(std::vector<double>{0.0, 0.0, 0.0})
.scan<'g', double>();
try {
program.parse_args(argc, argv); // Example: ./main --query_point 3.5 4.7 9.2
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
auto query_point = program.get<std::vector<double>>("--query_point"); // {3.5, 4.7, 9.2}You can also make a variable length list of arguments with the .nargs.
Below are some examples.
program.add_argument("--input_files")
.nargs(1, 3); // This accepts 1 to 3 arguments.Some useful patterns are defined like "?", "*", "+" of argparse in Python.
program.add_argument("--input_files")
.nargs(argparse::nargs_pattern::any); // "*" in Python. This accepts any number of arguments including 0.program.add_argument("--input_files")
.nargs(argparse::nargs_pattern::at_least_one); // "+" in Python. This accepts one or more number of arguments.program.add_argument("--input_files")
.nargs(argparse::nargs_pattern::optional); // "?" in Python. This accepts an argument optionally.Compound arguments are optional arguments that are combined and provided as a single argument. Example: ps -aux
argparse::ArgumentParser program("test");
program.add_argument("-a")
.default_value(false)
.implicit_value(true);
program.add_argument("-b")
.default_value(false)
.implicit_value(true);
program.add_argument("-c")
.nargs(2)
.default_value(std::vector<float>{0.0f, 0.0f})
.scan<'g', float>();
try {
program.parse_args(argc, argv); // Example: ./main -abc 1.95 2.47
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
auto a = program.get<bool>("-a"); // true
auto b = program.get<bool>("-b"); // true
auto c = program.get<std::vector<float>>("-c"); // {1.95, 2.47}
/// Some code that prints parsed arguments$ ./main -ac 3.14 2.718
a = true
b = false
c = {3.14, 2.718}
$ ./main -cb
a = false
b = true
c = {0.0, 0.0}Here's what's happening:
- We have three optional arguments
-a,-band-c. -aand-bare toggle arguments.-crequires 2 floating point numbers from the command-line.- argparse can handle compound arguments, e.g.,
-abcor-bacor-cab. This only works with short single-character argument names.-aand-bbecome true.- argv is further parsed to identify the inputs mapped to
-c. - If argparse cannot find any arguments to map to c, then c defaults to {0.0, 0.0} as defined by
.default_value
For inputs, users can express a primitive type for the value.
The .scan<Shape, T> method attempts to convert the incoming std::string to T following the Shape conversion specifier. An std::invalid_argument or std::range_error exception is thrown for errors.
program.add_argument("-x")
.scan<'d', int>();
program.add_argument("scale")
.scan<'g', double>();Shape specifies what the input "looks like", and the type template argument specifies the return value of the predefined action. Acceptable types are floating point (i.e float, double, long double) and integral (i.e. signed char, short, int, long, long long).
The grammar follows std::from_chars, but does not exactly duplicate it. For example, hexadecimal numbers may begin with 0x or 0X and numbers with a leading zero may be handled as octal values.
| Shape | interpretation |
|---|---|
| 'a' or 'A' | hexadecimal floating point |
| 'e' or 'E' | scientific notation (floating point) |
| 'f' or 'F' | fixed notation (floating point) |
| 'g' or 'G' | general form (either fixed or scientific) |
| 'd' | decimal |
| 'i' | std::from_chars grammar with base == 0 |
| 'o' | octal (unsigned) |
| 'u' | decimal (unsigned) |
| 'x' or 'X' | hexadecimal (unsigned) |
argparse provides predefined arguments and actions for -h/--help and -v/--version. These default actions exit the program after displaying a help or version message, respectively. These defaults arguments can be disabled during ArgumentParser creation so that you can handle these arguments in your own way. (Note that a program name and version must be included when choosing default arguments.)
argparse::ArgumentParser program("test", "1.0", default_arguments::none);
program.add_argument("-h", "--help")
.action([=](const std::string& s) {
std::cout << help().str();
})
.default_value(false)
.help("shows help message")
.implicit_value(true)
.nargs(0);The above code snippet outputs a help message and continues to run. It does not support a --version argument.
The default is default_arguments::all for included arguments. No default arguments will be added with default_arguments::none. default_arguments::help and default_arguments::version will individually add --help and --version.
argparse supports gathering "remaining" arguments at the end of the command, e.g., for use in a compiler:
$ compiler file1 file2 file3To enable this, simply create an argument and mark it as remaining. All remaining arguments passed to argparse are gathered here.
argparse::ArgumentParser program("compiler");
program.add_argument("files")
.remaining();
try {
program.parse_args(argc, argv);
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
try {
auto files = program.get<std::vector<std::string>>("files");
std::cout << files.size() << " files provided" << std::endl;
for (auto& file : files)
std::cout << file << std::endl;
} catch (std::logic_error& e) {
std::cout << "No files provided" << std::endl;
}When no arguments are provided:
$ ./compiler
No files providedand when multiple arguments are provided:
$ ./compiler foo.txt bar.txt baz.txt
3 files provided
foo.txt
bar.txt
baz.txtThe process of gathering remaining arguments plays nicely with optional arguments too:
argparse::ArgumentParser program("compiler");
program.add_arguments("-o")
.default_value(std::string("a.out"));
program.add_argument("files")
.remaining();
try {
program.parse_args(argc, argv);
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
auto output_filename = program.get<std::string>("-o");
std::cout << "Output filename: " << output_filename << std::endl;
try {
auto files = program.get<std::vector<std::string>>("files");
std::cout << files.size() << " files provided" << std::endl;
for (auto& file : files)
std::cout << file << std::endl;
} catch (std::logic_error& e) {
std::cout << "No files provided" << std::endl;
}
$ ./compiler -o main foo.cpp bar.cpp baz.cpp
Output filename: main
3 files provided
foo.cpp
bar.cpp
baz.cppNOTE: Remember to place all optional arguments BEFORE the remaining argument. If the optional argument is placed after the remaining arguments, it too will be deemed remaining:
$ ./compiler foo.cpp bar.cpp baz.cpp -o main
5 arguments provided
foo.cpp
bar.cpp
baz.cpp
-o
mainSometimes, several parsers share a common set of arguments. Rather than repeating the definitions of these arguments, a single parser with all the common arguments can be added as a parent to another ArgumentParser instance. The .add_parents method takes a list of ArgumentParser objects, collects all the positional and optional actions from them, and adds these actions to the ArgumentParser object being constructed:
argparse::ArgumentParser parent_parser("main");
parent_parser.add_argument("--parent")
.default_value(0)
.scan<'i', int>();
argparse::ArgumentParser foo_parser("foo");
foo_parser.add_argument("foo");
foo_parser.add_parents(parent_parser);
foo_parser.parse_args({ "./main", "--parent", "2", "XXX" }); // parent = 2, foo = XXX
argparse::ArgumentParser bar_parser("bar");
bar_parser.add_argument("--bar");
bar_parser.parse_args({ "./main", "--bar", "YYY" }); // bar = YYYNote You must fully initialize the parsers before passing them via .add_parents. If you change the parent parsers after the child parser, those changes will not be reflected in the child.
argparse::ArgumentParser program("json_test");
program.add_argument("config")
.action([](const std::string& value) {
// read a JSON file
std::ifstream stream(value);
nlohmann::json config_json;
stream >> config_json;
return config_json;
});
try {
program.parse_args({"./test", "config.json"});
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
nlohmann::json config = program.get<nlohmann::json>("config");argparse::ArgumentParser program("test");
program.add_argument("numbers")
.nargs(3)
.scan<'i', int>();
program.add_argument("-a")
.default_value(false)
.implicit_value(true);
program.add_argument("-b")
.default_value(false)
.implicit_value(true);
program.add_argument("-c")
.nargs(2)
.scan<'g', float>();
program.add_argument("--files")
.nargs(3);
try {
program.parse_args(argc, argv);
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
auto numbers = program.get<std::vector<int>>("numbers"); // {1, 2, 3}
auto a = program.get<bool>("-a"); // true
auto b = program.get<bool>("-b"); // true
auto c = program.get<std::vector<float>>("-c"); // {3.14f, 2.718f}
auto files = program.get<std::vector<std::string>>("--files"); // {"a.txt", "b.txt", "c.txt"}
/// Some code that prints parsed arguments$ ./main 1 2 3 -abc 3.14 2.718 --files a.txt b.txt c.txt
numbers = {1, 2, 3}
a = true
b = true
c = {3.14, 2.718}
files = {"a.txt", "b.txt", "c.txt"}argparse::ArgumentParser program("test");
program.add_argument("input")
.default_value("baz")
.action([](const std::string& value) {
static const std::vector<std::string> choices = { "foo", "bar", "baz" };
if (std::find(choices.begin(), choices.end(), value) != choices.end()) {
return value;
}
return std::string{ "baz" };
});
try {
program.parse_args(argc, argv);
}
catch (const std::runtime_error& err) {
std::cerr << err.what() << std::endl;
std::cerr << program;
std::exit(1);
}
auto input = program.get("input");
std::cout << input << std::endl;$ ./main fex
bazUse the latest argparse in your CMake project without copying any content.
cmake_minimum_required(VERSION 3.11)
PROJECT(myproject)
# fetch latest argparse
include(FetchContent)
FetchContent_Declare(
argparse
GIT_REPOSITORY https://github.com/p-ranav/argparse.git
)
FetchContent_MakeAvailable(argparse)
add_executable(myproject main.cpp)
target_link_libraries(myproject argparse)| Compiler | Standard Library | Test Environment |
|---|---|---|
| GCC >= 8.3.0 | libstdc++ | Ubuntu 18.04 |
| Clang >= 7.0.0 | libc++ | Xcode 10.2 |
| MSVC >= 14.16 | Microsoft STL | Visual Studio 2017 |
Contributions are welcome, have a look at the CONTRIBUTING.md document for more information.
Thanks goes to these wonderful people:
The project is available under the MIT license.